From 9fcf30f661e752e2c44df84286ae402349d8c4ff Mon Sep 17 00:00:00 2001
From: Claude Bot <claude-bot@bun.sh>
Date: Sun, 1 Feb 2026 00:30:16 +0000
Subject: [PATCH] fix(blob): handle UTF-8 paths correctly in stat() and
 delete() for Buffer paths

Use `ZigString.initUTF8()` instead of `ZigString.init()` when handling
non-encoded_slice PathLike variants. The latter doesn't mark bytes as UTF-8,
causing `toSliceClone()` to treat them as Latin1 and double-encode to UTF-8.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---
 src/bun.js/webcore/Blob.zig         |  2 +-
 src/bun.js/webcore/blob/Store.zig   |  2 +-
 test/regression/issue/26647.test.ts | 71 +++++++++++++++++++++++++++++
 3 files changed, 73 insertions(+), 2 deletions(-)
 create mode 100644 test/regression/issue/26647.test.ts

diff --git a/src/bun.js/webcore/Blob.zig b/src/bun.js/webcore/Blob.zig
index 28056fd2e0..63718d31fe 100644
--- a/src/bun.js/webcore/Blob.zig
+++ b/src/bun.js/webcore/Blob.zig
@@ -3134,7 +3134,7 @@ pub fn getStat(this: *Blob, globalThis: *jsc.JSGlobalObject, callback: *jsc.Call
                             .encoded_slice = switch (path_like) {
                                 // it's already converted to utf8
                                 .encoded_slice => |slice| try slice.toOwned(bun.default_allocator),
-                                else => try ZigString.init(path_like.slice()).toSliceClone(bun.default_allocator),
+                                else => try ZigString.initUTF8(path_like.slice()).toSliceClone(bun.default_allocator),
                             },
                         },
                     }, globalThis.bunVM());
diff --git a/src/bun.js/webcore/blob/Store.zig b/src/bun.js/webcore/blob/Store.zig
index 7de3d04c79..cc2b0e1fc0 100644
--- a/src/bun.js/webcore/blob/Store.zig
+++ b/src/bun.js/webcore/blob/Store.zig
@@ -263,7 +263,7 @@ pub const File = struct {
                 .path = .{
                     .encoded_slice = switch (path_like) {
                         .encoded_slice => |slice| try slice.toOwned(bun.default_allocator),
-                        else => try jsc.ZigString.init(path_like.slice()).toSliceClone(bun.default_allocator),
+                        else => try jsc.ZigString.initUTF8(path_like.slice()).toSliceClone(bun.default_allocator),
                     },
                 },
             }, globalThis.bunVM()),
diff --git a/test/regression/issue/26647.test.ts b/test/regression/issue/26647.test.ts
new file mode 100644
index 0000000000..3f49fe733a
--- /dev/null
+++ b/test/regression/issue/26647.test.ts
@@ -0,0 +1,71 @@
+import { expect, test } from "bun:test";
+import { tempDir } from "harness";
+
+// https://github.com/oven-sh/bun/issues/26647
+// Bun.file().stat() and Bun.file().delete() corrupt UTF-8 paths with non-ASCII
+// characters when the path is passed as a Buffer.
+
+test("Bun.file() with Buffer path handles UTF-8 correctly for stat()", async () => {
+  using dir = tempDir("test-26647", {
+    "über.txt": "content",
+  });
+
+  const filepath = `${dir}/über.txt`;
+
+  // Verify the file exists first using string path
+  const bunFile1 = Bun.file(filepath);
+  const stat1 = await bunFile1.stat();
+  expect(stat1.size).toBe(7); // "content" is 7 bytes
+
+  // Now test with Buffer path - this was failing before the fix
+  const bufPath = Buffer.from(filepath, "utf8");
+  const bunFile2 = Bun.file(bufPath);
+  const stat2 = await bunFile2.stat();
+  expect(stat2.size).toBe(7);
+});
+
+test("Bun.file() with Buffer path handles UTF-8 correctly for delete()", async () => {
+  using dir = tempDir("test-26647", {
+    "über.txt": "content",
+  });
+
+  const filepath = `${dir}/über.txt`;
+
+  // Test delete() with Buffer path - this was failing before the fix
+  const bufPath = Buffer.from(filepath, "utf8");
+  const bunFile = Bun.file(bufPath);
+
+  // Verify file exists before delete
+  const stat = await bunFile.stat();
+  expect(stat.size).toBe(7);
+
+  // Delete should succeed
+  await bunFile.delete();
+
+  // Verify file no longer exists
+  const exists = await Bun.file(filepath).exists();
+  expect(exists).toBe(false);
+});
+
+test("Bun.file() with Buffer path handles various UTF-8 characters", async () => {
+  using dir = tempDir("test-26647", {
+    "日本語.txt": "japanese",
+    "émoji🎉.txt": "emoji",
+    "中文测试.txt": "chinese",
+  });
+
+  // Test Japanese filename
+  const jpPath = Buffer.from(`${dir}/日本語.txt`, "utf8");
+  const jpStat = await Bun.file(jpPath).stat();
+  expect(jpStat.size).toBe(8); // "japanese" is 8 bytes
+
+  // Test emoji filename
+  const emojiPath = Buffer.from(`${dir}/émoji🎉.txt`, "utf8");
+  const emojiStat = await Bun.file(emojiPath).stat();
+  expect(emojiStat.size).toBe(5); // "emoji" is 5 bytes
+
+  // Test Chinese filename
+  const cnPath = Buffer.from(`${dir}/中文测试.txt`, "utf8");
+  const cnStat = await Bun.file(cnPath).stat();
+  expect(cnStat.size).toBe(7); // "chinese" is 7 bytes
+});